Scott wrote:
>I ask these question because I see as a key reason for the scientific
>failure of economics the acceptance by economists of virtually every
>stripe the need to restrict the problem spaces they specify to the
>limited domains of application of their particular techniques of
>analysis and associated agent representations.
It is not only in economics that this problem exists. I think it thrives
everywhere where people try to develop 'black box' models - those with
exogeneous variables - of complex systems in which all the variables are
essentially endogeneous. ie where you can't really say where the boundary
to the system is. This is of course a characteristic of almost all social
and economic systems.
I think that this 'black box' approach causes problems for most traditional
'engineering' modelling which seems only to be able to propose as practical
implementation solutions those things that the model captures. We see this
with traffic engineering in which the inability of current models to deal
with the full complexity of a streetgrid leads engineers to simplify the
geometry of their models until they are calculable and then to shut off the
'real' street connections that they cannot calculate so that the real world
mirrors their model.
The pity here is that instead of taking a 'scientific' approach to try and
understand the way the world works in all its complexity to the point at
which one has an explanatory and predictive theory (within their own
domain), the 'modelling' approach has tried to build models based on
theories borrowed from other domains (eg. 'gravity' applied to urban
systems) in the face of the evidence that suggests that these theories
don't hold in the new field of application. Of course, given a practical
problem and no better tools, there is nothing worong with this as an
engineering approach, just so long as one does not forget its status and
begin to think that the model somehow captures the way that the real system
works.
>In the absense of an intertwining ot these two strands, I see this
>scientific failure as a danger for the multiagent systems community in
>general and the social simulation community in particular.
I guess what Im saying is that the experience of the more general modelling
fraternity - which has a slightly longer history than MAS and soc sim -
bears out Scott's concerns. When the 'raison d'etre' for the community
becomes the model, science (at least the science of the application domain
if not the science of the model's logic and behaviour) seems to lose out.
Alan
________________________________________________________
Alan Penn
Director, VR Centre for the Built Environment
The Bartlett School of Architecture and Planning
1-19 Torrington Place (Room 335)
University College London, Gower Street, London WC1E 6BT
tel. (+44) (0)171 504 5919 fax. (+44) (0)171 916 1887
mobile. (+44) (0)411 696875
email. [log in to unmask]
www. http://www.vr.ucl.ac.uk/
________________________________________________________
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|